ANSYS#
ANSYS is a simulation software with a variety of functions including finite-element analysis and computational fluid dynamics.
Licensed Software
ANSYS is resticted to licensed users. Please contact help-rcc@mcw.edu with questions.
Multi-node Fluent job#
Fluent should be run from the ANSYS Workbench if possible, as this is the easiest method. However, if you need more resources than 1 node with 48 cores, you can run Fluent in a batch job on the cluster.
To get started, setup some file structure. Everytime we run a simulation, we should start with a new parent folder, which is example-job in this case. Make sure this is meaningful, since it will help you identify your simulations later.
cd /scratch/g/PI_NetID
mkdir -p example-job/{cas-files,output}
We created a new folder for our first simulation example-job in our group scratch directory. Within that folder, we also created a place to hold our input and output files. Upload the input cas files to the cas-files directory.
We need a journal file to tell Fluent how to proceed. Create a file /scratch/g/PI_NetID/example-job/fluent-test.jou.
/file/read-case-data "/scratch/g/PI_NetID/example-job/cas-files/test_fluent_case"
/solve/iterate 1000
/file/write-case "/scratch/g/PI_NetID/example-job/output/test_fluent_case.cas"
/file/write-data "/scratch/g/PI_NetID/example-job/output/test_fluent_case.data"
/parallel timer usage
/exit
yes
We also need a job script. Create a file /scratch/g/PI_NetID/example-job/job-script.slurm. Replace PI_NetID with your PI's username.
#!/bin/bash
#SBATCH --nodes=2
#SBATCH --ntasks-per-node=16
#SBATCH --job-name=test-fluent-case
#SBATCH --output=%x-%j.out
#SBATCH --time=168:00:00
# this is where we load software
module load impi
module load ansys/2022R1
export FLUENT_AFFINITY=0
export SLURM_ENABLED=1
export SCHEDULER_TIGHT_COUPLING=1
FL_SCHEDULER_HOST_FILE=slurm.${SLURM_JOB_ID}.hosts
/bin/rm -rf ${FL_SCHEDULER_HOST_FILE}
scontrol show hostnames "$SLURM_JOB_NODELIST" >> $FL_SCHEDULER_HOST_FILE
time fluent -g 3ddp -t${SLURM_NTASKS} -i fluent-test.jou -cnf=${FL_SCHEDULER_HOST_FILE}
Submit the job:
cd /scratch/g/PI_NetID/example-job
sbatch job-script.slurm
The example above is a job request for 2 nodes with 16 cores per node. Please make sure to insert your PI's username for PI_NetID.